I measured open vs closed models on my own product, and open weights won one and lost one

Описание и применение

Приложение Chalk использует открытые модели для генерации фото‑подтверждённых, точных котировок, повышая вероятность выигрыша и точность маржи.

Задача

A cleaning professional needed a quick, accurate, and data‑driven quote for a client visit, but manual quoting was slow, error‑prone, and often underpriced.

Результат

Open‑weight models produced more usable, accurate quotes than closed models, with higher win odds and better margin preservation.

Рабочий процесс

• Walk: capture photos and record job details
• Price: generate itemized scope, fair price, range, and win odds using open‑weight vision and TabPFN
• Send: compose client message with assumptions and validity window via open‑weight text model
• Answer: client accepts via link, booking the job and updating win rate

Доказательства

The vision model wrote the caption on each photograph — kitchen: greasy hob, tiled splashback, oven: caked‑on grime, interior racks — and attached each line item to the photograph that justifies it.

Источник

Two tasks, scored apart, so a weak pair of eyes can never be blamed on a weak brain.

Источник